Nanotransistors: A Bottom-Up View

Mark Lundstrom
Network for Computational Nanotechnology
Purdue University
West Lafayette, Indiana, USA
[email protected]

II.

AbstractThis talk describes how new understanding of

conduction at the molecular scale can inform nanoscale
MOSFET device engineering. A bottom-up approach
developed to explain conduction in molecules will be applied to
the analysis of nanotransistors. The generalization of the
simple approach to a rigorous treatment of quantum transport
with dissipation will be introduced. Some results on the
ultimate silicon MOSFET, the carbon nanotube MOSFET, and
the possibility of molecular transistors will be discussed. The
objective is to convey the essence of the approach and to
demonstrate its ability to simply and clearly described the
nanoscale MOSFET.

I.

INTRODUCTION

Silicon microelectronics has now become a true

nanoelectronic technology.
One definition of
nanotechnology involves structures that have at least two
dimensions below 100 nm, material properties that are sizedependent, and structures that include some aspect of selfassembly. Silicon nanoelectronics qualifies on all three
accounts and should be considered to be the first example of
active nanotechnology in manufacturing.
As silicon
technology is undergoing this transformation to
nanoelectronics, a whole new field of research has
developed. Unconventional nanoelectronics makes use of
semiconductor nanotubes and nanowires, and molecular
electronics uses molecules as electronic devices. It is not
clear yet where this research is heading in terms of practical
electronic technologies, but it does seem clear the research
will have impact on practical nanoelectronics in at least three
ways. The first is the development of new techniques for
metrology on the nanoscale. The second is the development
of new ways of assembling nanostructured materials and
nanoscale devices, and the third is a clear understanding of
electronic devices at the nanoscale. (Of course, we also hope
to discover new devices that can complement silicon CMOS
in system on chip applications.) In this talk, I will discuss
how the understanding of electronic conduction in molecules
can be applied to the nanoscale MOSFET. I hope to
demonstrate that this approach provides a fresh perspective
on the nanoscale MOSFET one that should prove useful as
technologists push the CMOS transistor to its limits.

1-4244-0301-4/06/$20.00 2006 IEEE.

33

MOLECULAR ELECTRONICS

The field of molecular electronics began in 1974 with the

prediction of Aviram and Ratner that molecules could be
designed to function as electronic devices [1]. By the
1990s, technology had been developed to allow the I-V
characteristics of individual molecules to be measured (see
[2, 3] for a review of the early work). As molecular
electronic technology has improved, the measurements have
become much more reproducible (e.g. [4]). For molecules
that do not undergo a change in conformation or oxidation
state, a general understanding of the measured I-Vs [5, 6]
and prescriptions to simulate them rigorously [7-14] have
been developed. My objective in this talk is to apply this
general understanding to the nanoscale MOSFET.
Fig. 1a is a schematic illustration of a typical experiment
[4] with the two electrodes being formed by a so-called break
junction and the molecule a small organic molecule [4]. Fig.
1b is a sketch of a typical I-V characteristic. Although the
result simply looks like two back-to-back diodes, there are
some interesting questions: 1) what controls the magnitude
of the conductance gap, 2) is conduction through the HOMO
(highest occupied molecular orbital, LUMO (lowest
unoccupied molecular orbitals, or both, and 3) how does one
explain the asymmetric I-V characteristics that are sometimes
observed? This simple molecular device is a useful place to
begin. By simply replacing one of the two metal contacts
with a semiconductor, for example, negative differential
resistance been observed [15, 16]. It can be readily
explained with this simple model, (although the origin of
negative differential resistance is still being debated [17]).
It is not possible to apply our conventional understanding
of electronic devices to molecular devices because concepts
such as bandstructure and effective mass have no meaning
for a molecule. It is possible, however, to develop a simple
theory for molecular devices based on the density-of-states,
which is always well defined and on the escape time for
electrons to leave the molecule through one of the two
contacts [18]. To begin, we shall assume that electron
transport is coherent (no scattering), but scattering processes
will also be included [18]. The two contacts are assumed to
be in thermodynamic equilibrium; a Fermi level
characterizes each one, EF1 and EF 2 = EF1 qV , where

contact 2

contact 1

Schematic illustration of a molecular device

Current

Fig. 1a

Voltage

conductance gap

Fig. 1b

Typical I-V characteristics of small organic molecules between two metallic electrodes

34

V is the applied bias. The device itself (the molecule) is

described by a density of states, D E U SCF , which
describes the broadened energy levels as shifted in energy
by the self-consistent electrostatic potential, U SCF . (This
simple approach ignores quantum interference between
levels, which allows us to deal with scalar quantities.) If
an electron is placed in the device, it eventually escapes
into contact 1 or contact 2 in a time, 1 or 2 . The
density of states and the escape times characterize the
molecular device there is no E(k), m*, or mobility at this
length scale.

nS = n1 + n2 = n1 1 + e qVDS / kB T cm -2 ,

(3)

where I have assumed Boltzmann statistics for simplicity.

The drain current of the MOSFET can be evaluated
from (2). We assume for simplicity, that the device is
large enough to be described by an E(k), further assume
parabolic energy bands with an effective mass, m*, and
write 1 = 2 = L x where L is the length of the

top of the barrier region and x is the ballistic velocity

directed along the channel. With these assumptions, we
can evaluate (2) to find

The number of electrons in the molecule is given by [5,

6, 18]

I D W = T (n1 n2 ) = T n1 1 e qVDS / kB T A / cm (4)

N = D1 (E U SCF ) f1 (E ) + D2 (E U SCF ) f2 (E ) dE (1)

energy

where D (E ) = D1 (E ) + D2 (E ) and D1,2 (E ) is the

density-of-states fillable by contact 1 or 2 [5, 6, 18].
Equation (1) is similar to the conventional expression for
the equilibrium carrier density in a semiconductor (e.g.
[19]) except that in this ballistic device we have two
different Fermi levels and states that are filled by one or
the other of these two Fermi levels.

L
(x)
position

The current through the device is given by [5, 6, 18]

ID = q

D(E)
( f (E) f2 (E))dE .
( 1 + 2 ) 1

Fig. 2 Energy band diagram of a MOSFET under high gate and drain
bias showing the critical, top-of-the-barrier region.

(2)

The important point to note is that current flows only

when there is a difference in the Fermi levels of the two
contacts. Each contact tries to fill the states in the device
according to its Fermi level. When the two Fermi levels
are different, there is a competition; one contact puts
electrons in the device and the other contact pulls them
out. When this happens, current flows. This simple model
explains a good deal about conduction through molecules
[5, 6, 18].
III.

where

is

the

T = 2kBT m

width

of

the

MOSFET

and

. Finally, we assume ideal, MOS

electrostatics to write

qnS = Cox (VGS VT ) .

(5)

Finally, by solving (3) (5), we obtain the I-V

characteristic of a ballistic transistor as

APPLICATION TO NANOSCALE MOSFETS

Fig. 2 shows the energy band diagram of a nanoscale

MOSFET under high gate a drain bias. The gate voltage
controls the height of the energy barrier between source
and drain. Well consider the small region near the top of
the barrier, which is electrostatically controlled by the gate
to be the device. If the region is long enough, it can be
characterized by a bandstructure, E(k), but even if it is not
large enough, it can always be described by a density of
states. The filling of these states according to the Fermi
levels of contact 1 (the source) and contact 2 (the drain)
can be evaluated from (1) and written as

I D = W CoxT

(1 e
(V V )
(1 + e
GS

qVDS / k B T
qVDS / k B T

).
)

(6)

Equation (6) is based on several assumptions. One is

Boltzmann statistics, which are not valid above threshold.
More generally, the exponentials should be replaced by
Fermi-Dirac integrals [20, 21] and the thermal velocity,
T by the so-called ballistic injection velocity,

35

 inj > T ,

Carriers cannot diffuse faster than the thermal velocity,

which depends on the location of the Fermi

I ON = W CoxT (VGS VT ) ,

thermal velocity, and then the current reaches the ballistic

limit. Equation (6) shows that once the mobility is high
enough, it no longer matters.
Finally, we should note that (8) is not a complete
model for a nanoscale MOSFET because the critical
length, L , has not been specified. For low VDS,
L L , the channel length, but for high drain bias, L
is essentially the distance over which the potential drops
by (kBT/q) or more generally, by [EF EC(0)]/q [22].

(7)

which is the traditional velocity saturation model except

that the ballistic injection velocity, a thermal equilibrium
quantity, replaces the saturated velocity. Equation (6)
implies that the velocity does saturate in a ballistic
MOSFET, but the physics has to do with how states in the
device are filled by the source and drain not by optical
phonon emission as in bulk silicon.

IV.

is

the

diffusion

coefficient.

Carriers cannot escape the device faster than the ballistic

escape time, so the escape time is the slower of the two
processes. The resulting I-V characteristic is

(
(

qV / k T
W Cox (VGS VT ) 1 e DS B
ID =
1
1 + e qVDS / kB T
1
+

Deff L
T

).
)

FULL QUANTUM DEVICE SIMULATION

The discussion in Sec. 3 provides a physically

intuitive, almost analytical treatment of the nanoscale
MOSFET, but detailed numerical simulations of realistic
structures are also needed. It turns out that our simple
conceptual picture of a nanoscale device can be
generalized into a rigorous treatment of quantum transport
[18], as illustrated in Fig. 3. All quantities become
matrices, which allows us to treat quantum interference
effects between levels through the off-diagonal matrix
elements.
For example, we describe the device by an
N

N
(
) Hamiltonian rather than by prescribing a
density of states. (Here N is the number of orbitals in an
atomistic treatment or the number of nodes in a finite
difference treatment.) The connections to the contacts are
described by the matrices, 1 and 2 and a self-

When (6) is compared to modern MOSFETs, we find

that present-day devices operate at roughly 50% of the
ballistic limit [23, 24]. Scattering is important and should
be included in a complete model. In our model, this is
easily accomplished by evaluating the escape times for
2
diffusive transport, 1 = 2 = L 2Deff , where

Deff = (k BT q )eff

L ) , exceeds the

so once the diffusion velocity, Deff

level [20 -23]. Equation (6) also assumes ideal, abovethreshold MOS electrostatics, but two-dimensional
electrostatics and subthreshold conduction can also be
treated [20 -23]. Equation (6) shows that the drain
saturation voltage of a ballistic MOSFET is a few kBT/q.
It also shows that the on-current is given by

(8)

[ ]

[ ]

[ ]

energy S describes incoherent scattering processes.

Instead of computing an electron density, we compute an

electron correlation function, G , an (N N )

matrix whose diagonal elements (in a real-space
representation) represent the position-dependent electron
density. Fig. 4 shows an example of such a calculation
the energy-resolved electron density inside a 10-nm
n

Referring to Fig. 2, the physical picture is that carriers

are injected from the source over the barrier where they
must diffuse across a short, low-field region of length, L.

36

devic
contact

[2]

[1]

Fig. 3

Fig. 4

contact

[H]. [S],

The non-equilibrium Greens function (NEGF) approach to quantum transport. (After Datta, [18]).

The energy-resolved electron density vs. position in a 10 nm, double gate MOSFET as computed by the
nanoMOS program [25].

37

a way to assess the performance potential of such a device

if the fabrication challenges could be overcome [32].
The results show that transistor behavior can be expected,
but that all of the traditional problems of the silicon
MOSFET are present. - two-dimensional electrostatics and
leakage currents due to tunneling. There are, however,
interesting possibilities. Molecules are flexible, and
dipoles might also be built in to respond differently to the
gate and drain fields [32]. This example shows how this
general technique can be used to explore and assess
potential device with a comm. conceptual and
computational framework.

channel length double gate MOSFET [25]. Although

quantum interference effects are apparent in the electron
density, the simulated I-V characteristics are very classical
[25]. The appearance of electrons under the potential
barrier is a consequence of quantum mechanical tunneling,
which degrades the subthreshold swing. Studies like these
are useful to explore the limits of scaling, and the same
method is readily applied to nanoscale MOSFETS, carbon
nanotube and semiconductor nanowire devices, and to
molecular devices. See Datta [18] for an introduction to
the method.
V.

BEYOND THE SILICON CMOS TRANSISTOR

VI.

The advantage of the bottom-up approach that I have

been describing is that it begins with a clear picture of
electronic conduction of the molecular scale and is not
based on concepts such as bandstructure, effective mass,
mobility, etc, which are macroscopic concepts that lose
validity at the nanoscale. The approach may be applied to
single molecule devices or to larger devices with
resolution at an atomistic scale. Two example will be
described in this section.

SUMMARY

In this talk, I presented and general conceptual

approach to nanoscale devices and showed how it provides
new insights into the performance of nanoscale

Because of their excellent electronic properties carbon

nanotube are of currently of great interest for electronic
device applications [26]. For many purposes, a single pz
orbital, tight-binding approach provides a reasonable
description of the electronic structure. Because of the
restricted phase space in one-dimension, it is possible to
treat transport at a much more rigorous level that is
possible for a conventional silicon MOSFET. The number
of atoms involves is reasonably small, so self-consistent
quantum ballistic transport is readily treated with atomistic
resolution [27]. Dissipative quantum transport can also be
handled by including the appropriate phonon modes and
computing the deformation potentials rigorously [28, 29].
Hot phonon effects can also be treated, either semiclassically [30] or in a full quantum approach [31]. Fig. 5
shows the energy-resolved current spectrum in a carbon
nanotube MOSFET under high gate and drain bias [29].
These results explain the puzzling observation that similar
devices appear to deliver nearly the current of a ballistic
device although strong electron-optical/zone boundary
phonon scattering occurs. As shown in Fig. 5, electrons
emit optical phonons and because the phonon energy is
large, electrons injected from the source into the channel
are unable to return to the source after emitting a phonon.
The result is that they must contribute to the DC drain
current, just as though no scattering occurred [29]. Of
course, the dynamic response will be affected, but only on
a very short timescale, and the increased build-up of
charge in the channel will affect the self-consistent
potential throughout the device, but these results explain
why strong phonon emission has only a small effect in
these devices.

EF

EF

Fig. 5

Energy-resolved current spectrum in a carbon nanotube MOSFET

as computed by the NEGF approach with electron-phonon
scattering included. (From [29])

MOSFETs. Also mentioned was the fact that the simple,

nearly analytical approach can be generalized into a
rigorous treatment of quantum transport in the presence of
dissipation. The result is a general conceptual and
computational framework to explore nanoscale electronic
devices. Extensions to spintronics are now underway [33].
One caveat must, however, be mentioned. The NEGF
approach is based on a single-particle, mean-field
approximation. Strongly correlated transport where effects
such as coulomb blockade occur are beyond the read of the
standard NEGF approach [18]. This topic is an area of
current research and may hold potential for novel
electronics devices. The main distinction is that Coulomb

The possibility of a molecular transistor has been of

great scientific and technological interest during the past
several years. The approach outlined in this talk provides

38

[7] P.S. Damle, A.W. Ghosh and S. Datta, "Firstprinciples analysis of molecular conduction using
quantum chemistry software, Chem. Phys. 281, 171188 (2002), Special Issue on Processes in Molecular
Wires.

Blockade shows charge quantization that usual quantum

transport misses, and this particle quantization requires a
special set of tools to handle properly [34, 35]. But it is
responsible for creating non-trivial IVs such as hysteretic
switching, random telegraph signals, single electron
charging, etc.

[8] A.W. Ghosh and S. Datta, "Molecular conduction:

paradigms and possibilities, J. Computational
Electronics, 1, 515, 2002.

In summary, the bottom-up approach presented here

provides a complementary perspective on nanoscale
electronics. I do not suggest that it replace traditional
engineering approaches based on drift-diffusion and
similar approaches, but I do encourage engineers to
become familiar with this alternative approach because it
provides a useful perspective that complements the
traditional one and new perspectives might lead us to new
device ideas. This paper presented the essence of the
technique. I hope that it motivates engineers to study the
references and add this conceptual and computational
framework to their engineering toolkit. Online lectures on
this topic are available at the nanoHUB [36].

[9] E. Emberly and G. Kirczenow, Theoretical study of

electrical conduction through a molecule connected to
metallic nanocontacts, Phys. Rev. B, 58, 1091110920 (1998).
[10] J. J. Palacios, E. Louis, A. J. Perez-Jimenez, E. San
Fabian and J. A. Verges, An ab-initio approach to
electrical
transport
in
molecular
devices,
Nanotechnology 13, 1-4, 2002.
[11] M. Di Ventra, S. T. Pantelides, and N. D. Lang,
First-principles calculation of transport properties of
a molecular device, Phys. Rev. Lett., 84, 979, 2000.

ACKNOWLEDGMENTS

[12] Mads Brandbyge, Jose-Luis Mozos, Pablo Ordejon,

Jeremy Taylor and Kurt Stokbro, Density functional
method for nonequilibrium electron transport, Phys.
Rev. B, 65, 16540, 2002.

The work described here took place over several years

and was supported by the Semiconductor Research
Corporation, The Microelectronics Advance Research
Corporation Focus Center for Materials, Structures, and
Devices, the Army Research Office Defense University
Research Initiative on nanotechnology, and the National
Science Foundation Network for Computational
Nanotechnology. The author thanks A.W. Ghosh for
illuminating discussions on molecular electronics.

[13] J. Taylor, H. Guo and J. Wang, Ab initio modeling of

quantum transport properties of molecular electronic
devices, Phys. Rev. B, 63, 245407, 2001.
[14] A. Pecchia, and A. DiCarlo, Atomistic theory of
transport in organic and inorganic nanostructures,
Reports on Progresses in Physics, 67, 1497, April,
2004.

REFERENCES
[1] T A. Aviram and M. A. Ratner, Molecular rectifiers.
Chem. Phys. Lett. 29, 277 (1974).
[2] A.W. Ghosh, P. Damle, S. Datta, and A. Nitzen,
Molecular electronics: Theory and device prospects,
MRS Bulletin, 29 (6), pp. 391-395, June, 2004.

[15] N. P. Guisinger, M. E. Greene, R. Basu, A. S. Baluch,

and M. C. Hersam, Room temperature negative
differential resistance through individual organic
molecules on silicon surfaces, Nano Lett. 4, 55 -59,
2004.

[3] C. Joachim, M.A. Ratner, Molecular electronics,

Proc. Nat. Acad. Sciences of the United States of
America, 102 (25), 880, June 21, 2005.

[16] Silicon-based molecular electronics, T. Rakshit, GC. Liang, A. W. Ghosh and S. Datta, Nano Lett. 4,
1803, 2004.

[4] J. Reichert, R. Ochs, D. Beckmann, H. B. Weber, M.

Mayor, and H. v. Lhneysen, Driving current
through single organic molecules, Phys. Rev. Lett.,
88, 176804, 2002.

[17] J. L. Pitters and R. A. Wolkow, Nano Letters,

Detailed studies of molecular conductance using
atomic resolution scanning tunneling microscopy,
Nano Lett., 6, 390 2006.

[5] M. Paulsson, F. Zahid and S. Datta, "Resistance of a

Molecule", chapter in Handbook of Nanotechnology,
ed. S. Lyshevski, CRC Press, 2003.

[18] S. Datta. Quantum Transport:

From Atom to
Transistor, Cambridge Univ. Press, Cambridge, UK,
2005.

[6] S. Datta "Electrical Resistance: An Atomistic View,

Nanotechnology, special issue on physics of
electronic transport in single atoms, molecules and
related nanostructures, eds. R.H. Friend and M.A.
Reed, May 2004 (IOP Publishing). (online tutorial :
Nanotechnology,15 , S433-S451 (2004).

[19] R.F. Pierret, Semiconductor Device Fundamentals,

Addison Wesley, Reading, MA, 1996.
[20] Anisur Rahman, Jing Guo, Supriyo Datta, and Mark
Lundstrom, Theory of ballistic nanotransistors,
IEEE Trans. Electron. Dev., 50, pp. 1853-1864, 2003.

39

[29] Siyuranga O. Koswatta , Sayed Hasan, and Mark S.

Lundstrom, M. P. Anantram, Dmitri E. Nikonov,
Ballisticity of nanotube FETs: role of phonon energy
and gate bias to appear in Appl. Phys. Lett.., 2006.

[21] Mark Lundstrom and Jing Guo, Nanoscale

Transistors: Physics, Modeling, and Simulation,
Springer, New York, 2005.
[22] Mark Lundstrom and Zhibin Ren, Essential physics
of carrier transport in nanoscale MOSFETs, IEEE
Trans. Electron Dev., 49, pp. 133-141, January, 2002.

[30] Sayed Hasan, Eric Pop, Ashraf Alam, Mark

Lundstrom, The role of self-heating and hot-phonons
in metallic single walled carbon nanotubes, condmat/06023662006.

[23] F. Assad, Z. Ren, D. Vasileska, S. Datta, and

M.S. Lundstrom, Performance limits of silicon
MOSFETS:
a theoretical study, IEEE Trans.
Electron Devices, 47, 232-240, 2000.

[31] L. Siddiqui, A.W. Ghosh, and S. Datta, Phonon

runaway in nanotube quantum dots, condmat/0606273, 2006.

[24] A. Lochtenfield and D.A. Antoniadis, On

experimental determination of carrier velocity in
deeply scaled NMOS: how close to the thermal
limit? IEEE Electron Dev. Lett., 22, pp. 95-97, Feb,
2001.

[32] A. W. Ghosh, T. Rakshit and S. Datta, Gating of a

molecular transistor electrostatic and conformational, Nano Lett. 4, 565, 2004.

[25] Zhibin Ren, Ramesh Venugopal, Sebastien Goasguen,

Supriyo Datta, and Mark S. Lundstrom nanoMOS
2.5: a two -dimensional simulator for quantum
transport in double-gate MOSFETs, IEEE Trans.
Electron. Dev., 50, pp. 1914-1925, 2003.

[33] S. Salahuddin, Prabhakar Srivastava, and S. Datta,

Integrating
spintronics
with
conventional
semiconductor devices through the exchange
interaction, presented at the 64th Device Research
Conference, The Pennsylvania State University,
University Park, Pennsylvania, June 26-28, 2006.

[26] P. McEuen, M. Fuhrer, and H. Park, Single walled

nanotube
electronics,
IEEE
Trans.
on
Nanotechnology, 1, pp. 78-85, 2002.

[34] B. Muralidharan, A. W. Ghosh and S. Datta, Probing

electronic excitations in molecular conduction, Phys.
Rev. B, 73, 155410, 2006.

[27] Jing Guo, Supriyo Datta, Mark Lundstrom, and

M.P. Anantram, Towards multi-scale modeling of
carbon nanotube transistors, Intl. J. on Multiscale
Computational Engineering, special issue on
multiscale methods for emerging technologies, N.
Aluru, editor, Vol. 2, pp. 257-276. 2004.

[35] K.M. Indlekoferm et al, Phys. Rev. B, 72, 125308,

2006.
[36] M.S. Lundstrom, Simple theory of the ballistic
nanotransistor, online lecture, www.nanoHUB.org.
2005.

[28] Siyu Koswatta, Mark S. Lundstrom, M.P. Anantram,

and Dmitri Nikonov, Simulation of phonon-assisted
band-to-band tunneling in carbon nanotube fieldeffect transistors, Appl. Phys. Lett., 87, 253107,
December 13, 2005.

40